Improved VMD method based on nonlinear Hoyer for abnormal impact interference

In Variational Mode Decomposition (VMD), kurtosis index is chosen for determining the optimal decomposition mode. However, the kurtosis is easily interfered by abnormal impact signal, which seriously affects the accuracy of VMD in fault extraction. To solve the above-mentioned problem, this paper proposes an improved VMD method based on nonlinear Hoyer (IVNH). Firstly, the whale optimization algorithm (WOA) is used to adaptively determine optimal decomposition parameters of VMD. Secondly, to weaken the effects of large amplitude impacts, the sigmoid function is applied to preprocess the decomposed modes. Thirdly, the Hoyer index with better sparsity, is applied to instead kurtosis to measure the sparse distribution of modes. Finally, calculate the Hoyer value of the decomposed modes, and the mode with the largest Hoyer will be chosen for signal reconstruction and envelope analysis. A set of bearing fault signals has been collected to verify the superiority of the proposed IVNH method. Results reveal that the proposed IVNH can effectively eliminate the interference of abnormal impacts.